BEGIN:VCALENDAR VERSION:2.0 PRODID:-//HasGeek//NONSGML Funnel//EN DESCRIPTION:AI agents don't behave like typical services: they have variab le control flow\, chained model and tool calls\, unpredictable latency and cost. X-WR-CALDESC:AI agents don't behave like typical services: they have varia ble control flow\, chained model and tool calls\, unpredictable latency an d cost. NAME:Platform engineering for production agents - a hands-on workshop X-WR-CALNAME:Platform engineering for production agents - a hands-on works hop REFRESH-INTERVAL;VALUE=DURATION:PT12H SUMMARY:Platform engineering for production agents - a hands-on workshop TIMEZONE-ID:Asia/Kolkata X-PUBLISHED-TTL:PT12H X-WR-TIMEZONE:Asia/Kolkata BEGIN:VEVENT SUMMARY:Platform engineering for production agents - a hands-on workshop DTSTART:20260919T083000Z DTEND:20260919T123000Z DTSTAMP:20261016T035215Z UID:session/WePQpubSxTVsdaDGsuxqe4@hasgeek.com SEQUENCE:16 CREATED:20260828T095012Z DESCRIPTION:# Platform Engineering for Production Agents\nA hands-on works hop for platform\, systems\, DevOps\, SRE\, and backend engineers who need to run AI agents safely\, reliably\, and cost-effectively in production.\ n\nAI agents are no longer just local coding assistants or experimental ch atbots. Once agents call tools\, access data\, run workflows\, retry faile d steps\, and serve users\, they become production workloads. They need th e same platform discipline as any other distributed system: identity\, sec rets\, isolation\, observability\, deployment\, rollback\, cost controls\, and operational recovery.\n\nThis workshop focuses on the runtime and pla tform responsibilities that come with supporting agent-based applications in production.\n\n## What is this workshop about?\nThis workshop will help engineers understand what changes when AI agents move from development en vironments into hosted\, production systems.\n\nParticipants will learn ho w an agent request flows across model calls\, tool calls\, retries\, respo nses\, logs\, traces\, and failure states. The workshop will cover the ope rational controls needed to make agents observable\, bounded\, secure\, de buggable\, and recoverable.\n\nThe workshop is not about prompt engineerin g or chatbot design. It is about the production environment around agents: how to run them\, monitor them\, constrain them\, and recover them when s omething goes wrong.\n\n## What will participants work on?\nParticipants w ill work with a mock production incident-management agent built with pi.de v. The agent will run against a simple Go application connected to PgBounc er\, PostgreSQL\, and Redis. The system emits metrics and traces\, and inc ludes routes that fail in predictable ways.\n\nThrough this setup\, partic ipants will examine how agent workloads behave in production-like conditio ns\, including:\n\n- Liveness and readiness failures: is the agent working \, and can it continue accepting requests?\n- Observability gaps: what is the agent doing when it appears stuck?\n- Cost failures: how can runaway t oken or tool usage be detected and controlled?\n- Safety failures: how can agents be prevented from taking unsafe actions?\n- Correctness failures: how do we evaluate whether an agent's completed work was actually right?\n \nParticipants will be introduced to the sample system\, the pi.dev incide nt-management agent\, and a sequence of failures that help frame the platf orm engineering controls required for production agents.\n\n## Who is the target audience?\nThis workshop is for:\n\n- Platform engineers supporting application teams that are building or deploying AI workloads\n- Systems engineers working on infrastructure\, runtime environments\, internal plat forms\, or developer platforms\n- DevOps and SRE practitioners responsible for reliability\, deployment\, observability\, security\, and cost contro ls\n- Backend engineers building agents who want to understand the product ion platform their agents will need\n- Engineers experimenting with AI age nts who understand containerisation basics and want to move beyond local d emos\n\n## Background knowledge requirements\nYou do not need prior experi ence building AI agents.\n\nYou should be comfortable with basic container ization concepts and familiar with how production applications are usually deployed\, monitored\, and operated.\n\nThis workshop is not designed for non-technical participants\, business users\, or people looking only for an introduction to prompting or AI tools.\n\n## Workshop content and plan\ nThe workshop will be hands-on and will cover the following areas:\n\n### 1. What makes agents different as production workloads\n- Difference betwe en coding assistants\, local agents\, and hosted agent services\n- What mo dels\, harnesses\, and agents are\n- Anatomy of an agent request\n- Model calls\, tool calls\, retries\, memory\, state\, and responses\n- Why agent workloads are harder to reason about than standard request-response servi ces\n\n### 2. Runtime behaviour and failure modes\n- Tracing an agent requ est across model and tool boundaries\n- Latency\, timeout\, retry\, and pa rtial failure patterns\n- Liveness\, readiness\, observability\, cost\, sa fety\, and correctness failures\n- Model failures\, tool failures\, permis sion failures\, and workflow failures\n- Choosing safe recovery actions wh en agent execution goes wrong\n\n### 3. Observability for agents\n- Logs\, metrics\, and traces for agent systems\n- Tracking token usage\, latency\ , errors\, tool calls\, and cost\n- Debugging agent behaviour using teleme try\n- Separating application bugs\, model issues\, and platform failures\ n\n### 4. Access\, isolation\, and safety boundaries\n- Bounding what an a gent can access\n- Tool permissions and scoped credentials\n- Secrets mana gement for agent workloads\n- Sandboxing\, isolation\, rate limits\, time limits\, and spend limits\n\n### 5. Platform responsibilities\n- What belo ngs in agent code vs what should be provided by the platform\n- Minimum pr oduction environment for hosted agents\n- Deployment\, rollback\, configur ation\, and environment management\n- Operational checklists for supportin g agent-based applications\n\n### 6. Hands-on exercises\nParticipants will work with a small agent workload and practice:\n\n- Running it in a conta inerised environment\n- Tracing a request through model calls and tool cal ls\n- Observing latency\, failures\, token usage\, and cost\n- Applying ru ntime limits and access boundaries\n- Diagnosing failures and selecting re covery actions\n- Evaluating correctness and performance over time\n\n## W orkshop agenda\n- 14:00-14:25: Introduction\, what is an agent\, and why a gent workloads differ from traditional workloads\n- 14:25-15:00: Defining the system under test and the test agent\; explaining the pi.dev agent and its tools\n- 15:00-15:30: Failure modes in agents\, and how these differ from failures in traditional applications\n- 15:30-16:00: Networking and c offee break\n- 16:00-16:20: Observability\, liveness\, and readiness for a gents as production applications\n- 16:20-17:00: Infrastructure\, cost con trol\, and security boundaries\n- 17:00-17:15: Evaluating correctness and performance over time\n- 17:15-17:30: A unified layer for addressing agent failure modes\n- 17:30-18:00: Open Q&A and wrap-up\n\n## Key takeaways\nB y the end of the workshop\, you will be able to:\n- Explain how production agents differ from local coding tools and simple AI demos\n- Trace one ag ent request across model calls\, tool calls\, retries\, and responses\n- U se metrics and traces to identify latency\, errors\, token usage\, and cos t\n- Bound an agent's access to tools\, data\, time\, credentials\, and sp ending\n- Decide which controls belong in agent code and which should be p latform defaults\n- Define the minimum production environment for identity \, secrets\, isolation\, telemetry\, deployment\, and rollback\n- Diagnose model\, tool\, and agent failures and choose a safe recovery action\n- Ev aluate correctness and performance for an agent workload over time\n- Supp ort application teams that are moving agent workloads into production\n\n# # Software installation requirements\nParticipants will need:\n- A laptop that can comfortably run 2-3 Docker containers\n- Linux or macOS preferred \n- Docker installed before the workshop\n- An API key for any model provi der supported by pi.dev\n- A code editor\n- Basic familiarity with command -line usage\n\nParticipants may use any LLM API provider that is compatibl e with the workshop exercises.\n\n**Low-cost or free options include:**\n- OpenCode: create an API key on OpenCode Zen and use the free `big-pickle` model. This model may be rate limited.\n- OpenRouter: create an account\, generate an API key\, and use one of the free models listed on OpenRouter \, such as `nvidia/nemotron-3.5-lightning:free`.\n\nDetailed setup instruc tions will be shared with registered participants before the workshop. Reg istered participants may also receive a test version of the workshop code before the session so they can check their setup.\n\n## Workshop code\nPar ticipants will work through code from: https://github.com/ace03uec/platfor m_engineering_production_agents\n\nWorkshop Presentation : https://drive.g oogle.com/file/d/1khDCpBL_bVF39cKaOFAOfPD2rNnp2Lrl/view?usp=sharing\n\n## About the instructor\nGiridharan Manivannan is a DevOps and Platform Engin eering consultant with over 12 years of experience. He has worked with inf rastructure teams at Razorpay and Hasura\, helping teams build\, operate\, and scale production systems.\n\nHe writes and shares thoughts on DevOps\ , platform engineering\, AI infrastructure\, Chelsea FC\, and anime at htt ps://x.com/_dforce. He can be reached at hello@giridharan.dev for work.\n\ n## How to attend this workshop\nThis workshop is open to Rootconf members :\n\nhttps://hasgeek.com/rootconf#memberships\n\nThe workshop is open to 3 0 in-person participants\, with hybrid access available for remote attende es. Seats for in-person participants will be available on a first-come\, f irst-served basis.\n\n## Contact information\nFor inquiries about the work shop\, contact +91-7676332020 or write to info@hasgeek.com. LAST-MODIFIED:20260919T082750Z LOCATION:Bangalore - https://hasgeek.com/rootconf/platform-engineering-for -production-agents/ ORGANIZER;CN=Rootconf:MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/rootconf/platform-engineering-for-production-agent s/ BEGIN:VALARM ACTION:display DESCRIPTION:Platform engineering for production agents - a hands-on worksh op in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT END:VCALENDAR